Drones and AI Take On Killer Sharks Down Under

Robots are an expensive but cool solution to a very rare problem

3 min read

Whether or not shark attacks are a major problem in Australia (spoiler alert: they're not), the Australian government has devoted an enormous amount of resources into trying to mitigate the risk of sharks near popular beaches. They've tried nets to keep the sharks out, they've tried electronic gadgets to dissuade them, and they've tried lots of different ways of killing them, without much in the way of evidence that any of it is particularly effective. 

After six months of trials, the latest and most robot-y idea is about to be implemented: drones will start patrolling some Australian beaches next month, using cameras and some AI-backed image analysis software to spot lurking sharks much better than humans can.

Humans aren't particularly good at identifying sharks on aerial imagery. We can manage a 20-30 percent accuracy rate, which means both identifying other things as sharks (kinda bad) and misidentifying sharks as other things (way worse). As with many tasks of this kind, a machine learning system does much better: once it's been trained on labeled aerial videos of sharks, whales, dolphins, surfers, swimmers, boats, and whatever else, the software is 90 percent accurate at telling humans to panic because there's a shark somewhere. And when implemented on a drone, the system really does tell people to panic, using a loudspeaker to warn them that there's a shark in the water. 

The drones come from an Australian company called Westpac Little Ripper, which modifies a few different kinds of commercial drones for tasks like shark spotting as well as general lifesaving operations, such as dropping beacons and even rafts. The larger Little Ripper drones are gas powered and can fly for hours, which is nice, but they somehow cost up to US $250,000 each.

It's also worth noting that visual shark detection from the air only works when the weather's good, and while you can use it to reliably spot sharks near the surface, ones that are deeper down (hunting you from below) are still very hard to spot.

In case you were wondering whether any of this is driven by politics rather than common sense, the mayor of one Australian beach town would like you to know that “the Westpac Little Ripper Lifesaver have been developing methods to identify sharks in the water since its inception, and the latest news is groundbreaking. It is a major step towards solving the problem of shark attacks, and I am delighted to support safety in the water.”

Problem is, it's not at all clear that this is a major step, or even a minor step, towards solving the problem of shark attacks, and it may not make the water all that much safer. Besides the issues with aerial shark spotting as mentioned above, there's the simple fact that shark attacks are extraordinarily rare. In 2016, there were 26 unprovoked shark attacks in Australia, resulting in 16 injuries and 2 fatalities. To put that in context, in 2016, about 120 people drowned on the Australian coast, and there were also 1,290 road fatalities. In other words, you should be much more worried about many things that aren't sharks, like driving to and from the beach and getting out of the water safely. Here's two more: you are about four times more likely to be struck by lightning than bitten by a shark in Australia, and you are also more likely to drown in your own bathtub than be killed by an Australian shark. Oh, and while we're at it, humans kill something like 100 million sharks every year, just for a little perspective on the side of the shark.

Still, I suppose there's no reason why we shouldn't try to mitigate shark attacks whenever possible, even if the amount of money and effort going into preventing something that doesn't happen very often at all might seem a bit much. So sure, if it makes people feel better, send up those shark spotting drones, but the ocean is a dangerous place, and all things considered, sharks are almost certainly not the most dangerous thing out there. These drones would offer a lot more value if we focused on the fact that they can detect people and drop lifesaving equipment to them much faster and safer than a human lifeguard could, but that’s not nearly as exciting as trying to take on sharks.

[ Reuters ]

The Conversation (0)

How the U.S. Army Is Turning Robots Into Team Players

Engineers battle the limits of deep learning for battlefield bots

11 min read
Robot with threads near a fallen branch

RoMan, the Army Research Laboratory's robotic manipulator, considers the best way to grasp and move a tree branch at the Adelphi Laboratory Center, in Maryland.

Evan Ackerman

“I should probably not be standing this close," I think to myself, as the robot slowly approaches a large tree branch on the floor in front of me. It's not the size of the branch that makes me nervous—it's that the robot is operating autonomously, and that while I know what it's supposed to do, I'm not entirely sure what it will do. If everything works the way the roboticists at the U.S. Army Research Laboratory (ARL) in Adelphi, Md., expect, the robot will identify the branch, grasp it, and drag it out of the way. These folks know what they're doing, but I've spent enough time around robots that I take a small step backwards anyway.

This article is part of our special report on AI, “The Great AI Reckoning.”

The robot, named RoMan, for Robotic Manipulator, is about the size of a large lawn mower, with a tracked base that helps it handle most kinds of terrain. At the front, it has a squat torso equipped with cameras and depth sensors, as well as a pair of arms that were harvested from a prototype disaster-response robot originally developed at NASA's Jet Propulsion Laboratory for a DARPA robotics competition. RoMan's job today is roadway clearing, a multistep task that ARL wants the robot to complete as autonomously as possible. Instead of instructing the robot to grasp specific objects in specific ways and move them to specific places, the operators tell RoMan to "go clear a path." It's then up to the robot to make all the decisions necessary to achieve that objective.

Keep Reading ↓ Show less